On 2012-02-01 11:13:54 -0500, Justin Ruggles wrote: > On 01/31/2012 08:44 PM, Janne Grunau wrote: > > > On 2012-02-01 02:21:35 +0100, Janne Grunau wrote: > >> On 2012-01-31 11:40:16 -0500, Justin Ruggles wrote: > >>> av_realloc() does not guarantee alignment, which is required for > >>> DSPContext.bswap_buf(). > >>> --- > >>> libavcodec/apedec.c | 9 +++++---- > >>> 1 files changed, 5 insertions(+), 4 deletions(-) > >>> > >>> diff --git a/libavcodec/apedec.c b/libavcodec/apedec.c > >>> index fa50d61..71cd6ef 100644 > >>> --- a/libavcodec/apedec.c > >>> +++ b/libavcodec/apedec.c > > > > [...] > >>> @@ -838,10 +840,9 @@ static int ape_decode_frame(AVCodecContext *avctx, > >>> void *data, > >>> return AVERROR_INVALIDDATA; > >>> } > >>> > >>> - tmp_data = av_realloc(s->data, FFALIGN(buf_size, 4)); > >>> - if (!tmp_data) > >>> + av_fast_malloc(&s->data, &s->data_size, FFALIGN(buf_size, 4)); > >>> + if (!s->data) > >>> return AVERROR(ENOMEM); > >>> - s->data = tmp_data; > >>> s->dsp.bswap_buf((uint32_t*)s->data, (const uint32_t*)buf, > >>> buf_size >> 2); > >>> s->ptr = s->data; > >>> s->data_end = s->data + buf_size; > >> > >> ok > > > > actually not unless buf_size is guaranteed to be a multiple of 4. > > > I see what you mean, but that would really be a separate issue in the > current code. Probably it should be: > s->data_end = s->data + (buf_size & ~3);
that and there is a gap between the padding and the last bswapped content if buf_size is not a multiple of 4. Janne _______________________________________________ libav-devel mailing list [email protected] https://lists.libav.org/mailman/listinfo/libav-devel
